Hassan Bevrani

Professor, University of Kurdistan

Fall 2023

H. Bevrani University of Kurdistan 1

Contents

1. Nominal Plant, Uncertainty (Spinning Satellite)

2. Performance Weight
3. Generalized Plant (Interconnection)
4. Optimal Control Design
5. Closed-Loop Analysis
6. H2 Controller

H. Bevrani University of Kurdistan 2

 Reference

1. S. Skogestadand I. Postlethwaite, Multivariable Feedback Control; Analysis and

Design, Second Edition, Wiley, 2005.
2. M. Fujita, Lecture Notes on Feedback Control Systems, Tokyo Institute of
Technology, 2019.
3. R. Smith, Lecture Notes on Control Systems, ETH Zurich, 2020.

H. Bevrani University of Kurdistan 3

Nominal Plant Model

o Transfer
Function Matrix

o State Space Representation

H. Bevrani University of Kurdistan 4

 Characteristics of Nominal Plant Model

o Poles:
Unstable Poles
Vibrational System

o Frequency Response ( )

H. Bevrani University of Kurdistan 5

Plant Model with Output Uncertainty

o Uncertain Plant Model

Gain Margin:
Delay Margin:

o Multiplicative (Output) Uncertainty

Uncertainty Weight:

H. Bevrani University of Kurdistan 6

 Uncertainty Weight

H. Bevrani University of Kurdistan 7

Update Uncertainty Weight

Uncertain Factors =
Gain, Time Delay＋ Unmodeled Dynamics (High Freq.)

Target Loop (Roll-off)

H. Bevrani University of Kurdistan 8

 Time Responses for Uncertain Plant Model

H. Bevrani University of Kurdistan 9

Performance Weight

H. Bevrani University of Kurdistan 10

 Update Performance Weight

H. Bevrani University of Kurdistan 11

Control Problem Formulation

Nominal Performance Robust Stability

H. Bevrani University of Kurdistan 12

 Continue

Generalized Plant (Interconnection building):

Mixed Sensitivity

H. Bevrani University of Kurdistan 13

H∞ Optimal Controller
H∞ Control Problem

Given , find all stabilizing controllers K such

that .

[K, CL, gam, info] = hinfsyn(G, nmeas, ncon, key1, value1, key2, value2, …)
Input arguments Output arguments
G: generalized plant K: LTI controller
nmeas: number of control inputs CL: closed loop system which consists of and
ncon: number of measurement outputs gam: H∞ norm of closed loop system
info: information of output results

key settings
Gmax: upper limit of gam(=Inf)
Gmin: lower limit of gam(=0)
Tolgam: relative error of gam(=0.01) Method
So: frequency at which entropy is assessed (default=Inf)
ric: Ricatti solution (default)
Display: off (not show setting process)
lmi: LMI solution
on(show setting process)
maxe: max entropy solution
H. Bevrani University of Kurdistan 14
 H∞ Optimal Controller

o Assumptions for Generalized Plant

“If the Robust Control toolbox of MATLAB complains, then it probably means
that your control problem is not well formulated, and you should think again”

H. Bevrani University of Kurdistan 15

-iteration to obtain H∞ Controller

o Check 1 Appropriately sub-optimal (Default settings)

Closed-loop Transfer Function

o Check 2

0.5);

(Ref 1, p. 358) “interested” frequency range

H. Bevrani University of Kurdistan 16
 H∞ Controller

Order: 6

Poles:
Mixed Sensitivity Problem
⇒Pole/Zero Cancellations

Zeros: H∞ Loop-shaping Design

H. Bevrani University of Kurdistan 17

Open-loop Frequency Response Analysis

(corresponding maximum stability margin)

Loop Transfer Function

H. Bevrani University of Kurdistan 18

 Continue

Nominal Performance (NP) Test Robust Stability (RS) Test

√ √
H. Bevrani University of Kurdistan 19

Closed-loop Performance Analysis

o Nominal Stability (NS)

Poles of

H. Bevrani University of Kurdistan 20

 Continue

Nominal Performance (NP) Robust Stability (RS)

H. Bevrani University of Kurdistan 21

Closed-loop Systems Time Responses Analysis

Automatic Selection
Manual Selection

H. Bevrani University of Kurdistan 22

 H∞ Controller Using mixsyn function
[K, CL, gam, info] = mixsyn(P, WP, WU, WM, key1, value1, …)

Remark:

Perfect matching

H. Bevrani University of Kurdistan 23

Important Note in MIMO Loop Shaping

SISO Loop Shaping MIMO Loop Shaping

H. Bevrani University of Kurdistan 24

 Infeasible Performance Weight

√ X
Controller exists No Controller

H. Bevrani University of Kurdistan 25

Important Note: Ensuring Assumptions

o Generalized Plant

o Example

H. Bevrani University of Kurdistan 26

 Example: H2 Controller
[K, CL, gam, info] = h2syn(G, nmeas, ncon)

Remark:
Error!

H. Bevrani University of Kurdistan 27

Project: Report 7

Consider your dynamic system :

Using uncertainty and performance weighing functions (W2 and W1)
in the previous project steps, Design an H∞ controller.

Deadline: The day before next Meeting

Please only use this email address: [email protected]

H. Bevrani University of Kurdistan 28

 Thank You!

H. Bevrani University of Kurdistan 29